K means method and LBP based methods are commonly used for manipulating digital wave files, and they can be used for wave file features extraction, these features can be used later on as a wave file key or signature which refers to the wave file in any application, such as voice recognition or voice retrieval applications. K means method can be used to extract voice features, these features can be represented by clusters points, or within_cluster sums, or by the clusters centroids, and in this paper we will investigate the both methods in order to do some recommendations regarding the efficient method of voice features extraction.